DACE Based Probabilistic Optimization of Mechanical Components

Abstract

Application of DACE (Design and Analysis of Computer Experiments) methods for probabilistic design space exploration and optimization to the design of a mechanical component is demonstrated. The key part of the paper is focused on the problem formulation and process flow for performing a probabilistic optimization. The authors have shown that for computationally intensive problems, probabilistic optimization can be carried out efficiently within a DACE framework. For problems that are not costly to compute, direct probabilistic optimization can be carried out by the efficient integration of probabilistic analysis and global optimization (such as Genetic Algorithms). The strategy in the paper proves to be especially beneficial for those organizations that are reluctant to move to probabilistic methods and also for the current practitioners of probabilistics. The methodology is illustrated with examples from both simple and computationally intensive engineering problems.